Rosetta GAdock
Rosetta (pre-release version, GIT hash 845e653c80ec0965c63afa747b6c7efde2ef63b8)/CORINA Classic, webserver version/OpenBABEL-2.4.1/Antechamber-17.3
Turned on --amide_chi option at Rosetta parameter generation stage to allow trans/cis flipping of amide bond dihedrals
AM1BCC partial charge
1) Converted SMILES into hydrogen-attached mol2 file using Corina. (NOTE: for several ligands OpenBABEL was used instead due to failures in Corina)
2) Partial charge and optimal bond geometry of the ligands calculated using Antechamber.
3) Rosetta atom typing was assigned using mol2genparams.py from Rosetta
4) Receptor structure brought from PDBIDs 3dv1 and 3dv5, ran Rosetta FastRelax with options
"-constrain_relax_to_start_coords -ramp_constraints false –beta_cart_nov16 -dualspace".
Relax was run with the native ligands present in both cases.
All three chains in the asymmetric unit were individually relaxed and the lowest-energy configuration of each was used in docking.
Rosetta Generalized Energy function; beta_genpot.wts in the repository
10 iterations and pool size of 100 in Rosetta GAdock, 10 repeats of docking per ligand
Soften energy function at grid docking stage, LJ repulsion weight at 35%
Standard energy function at final relax stage
See attached XML script and command line in attached SuppInfo file.
Rosetta’s GALigandDock, a new genetic-algorithm (GA)-based ligand docking method was used.
An overview:
(1) “LigandAligner” in Rosetta GAdock module was used to prepare 40% of the initial population
by matching ligand to a set of reference interaction motifs from PDB IDs 3dv1 and 3dv5;
the remaining 60% of inputs were generated by randomization around the binding pocket.
(2) Starting from this pool, Rosetta GALigandDock runs a GA optimization guided by the Rosetta generalized energy function.
For efficiency, energies are precomputed on a 0.333Å grid and spline interpolation is used
to quickly evaluate ligand energetics. Flexible sidechains are assigned at this point
by considering distances and orientations of neighboring CB atoms to the center of ligand;
on average sidechains of 10 residues are allowed to move in this challenge.
(3) At every iteration of GA, new structures are generated by mutation/crossover over internal coordinates of all flexible DOF.
For each ligand conformation considered, grid-based energy optimization is performed:
discrete sidechain rotamers are sampled, then all pocket sidechains and ligands are minimized in internal coordinates.
(4) At the end of every iteration, 100 new structures are combined with 100 original structures,
then 100 structures with lowest energy - and not closer to any lower energy structure than 2.0Å RMS -
are selected for the next round.
(5) After 10 iterations of GA, the final structures are relaxed with explicit energy calculation (i.e. not using grid);
in this stage flexible sidechains and backbone atoms are minimized in Cartesian space,
applying coordinate restraints to all flexible DOFs with a harmonic constant of 0.1 kcal/mol.
The pose with the best total energy after 10 repeats of docking was selected as model 1.
No
Yes